HFBoards

Go Back   HFBoards > General Hockey Discussion > By The Numbers
Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
By The Numbers Hockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.

Post-consolidation VsX Benchmarks

Reply
 
Thread Tools
Old
02-27-2013, 05:40 AM
  #1
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Post-consolidation VsX Benchmarks

Allright, I have done the hard work of going through every post-consolidation NHL season and trying to set some kind of benchmark against which we can compare scoring in a VsX percentages system. Before I post the results, my methodology:

1. First preference is to use the #2 scorer

2. If #3 points/#2 points < .90, I use the #3 scorer, unless...

3. There is a gap of greater than 10% anywhere else in the top-5 - following the same method as above: [small #]/[large #] < .90.

At that point, I take the first gap, and identify the upper outlier group (top 3 or 4 or 5 above which the gap occurs), and then go down into the scoring table until I reach a number of players which equals: [size of outlier group] * 2. The benchmark is set as an average of the scoring of these players.

4. If any player in the top-5 is more than 7% below the player above him and more than 7% above the player below him, his score is taken as the benchmark. [this is the Bathgate Rule]

5. The Wartime Fudge, as discussed in posts 131 and 159 of this thread. The new benchmark numbers are shows in parentheses next to the originals.

Examples of the methodology:

1. 2003-04:

Quote:
The top of the scoring table is:

1. St. Louis - 94
2. Sakic - 87
2. Kovalchuk - 87
4. Naslund - 84
5. Hossa - 82
6. Elias - 81

Joe Sakic and Ilya Kovalchuk are tied in second place on the scorer's list with 87 points. There are no gaps of greater than 10% in the rest of the top-5, so Sakic/Kovalchuk, as co-#2 scorers, are used as the benchmark.
2. 2005-06:

Quote:
The top of the scoring table is:

1.Thornton - 125
2. Jagr - 123
3. Ovechkin - 106
4. Heatley - 103
4. Alfredsson - 103
6. Crosby 102

There is a 14% gap between Jagr's 123 and Ovechkin's 106 points, but no other large gaps in the top-5, so Ovechkin's 106 points as the #3 scorer is used as the benchmark.
3. 1979-80:

Quote:
The top of the scoring table is:

1. Dionne - 137
1. Gretzky - 137
3. Lafleur - 125
4. Perreault - 106
5. Rogers - 105
6. Trottier - 104

There is only a 9% gap between the #2 and #3 scorers here, but there is a 15% gap between the #3 and #4 scorers. The outlier group is the top three, so we average the scoring of the top 6 players to set our benchmark, which ends up being 119 points - a completely artificial number.
4. 1956-57:

Quote:
1. Howe - 89
2. Lindsay - 85
3. Beliveau - 84
4. Bathgate - 77
5. Litzenberger - 64
...average as benchmark: 77

Here, Andy Bathgate is more than 7% behind Beliveau above him, and more than 7% ahead of Litzenberger below him. Bathgate's score is taken as the benchmark in this season.
Ok, without further ado, here is the data:

Quote:
1926-27:
1. Cook - 37
2. Irvin - 36
3. Morenz - 32
...benchmark = Morenz: 32

1927-28:
1. Morenz - 51
2. Joliat - 39
3. Boucher - 35
3. Hay - 35
...benchmark = Boucher: 35

1928-29:
2. Stewart - 29

1929-30:
2. Boucher - 62

1930-31:
1. Morenz - 51
2. Goodfellow - 48
3. Conacher - 43
...benchmark = Conacher: 43

1931-32:
2. Primeau - 50

1932-33:
2. Jackson - 44

1933-34:
1. Conacher - 52
2. Primeau - 46
3. Boucher - 44
4. Barry - 39
4. Dillon - 39
4. Stewart - 39
...average as benchmark: 43

1934-35:
2. S. Howe - 47

1935-36:
2. Thompson/Barry - 40

1936-37:
2. Apps - 45

1937-38:
1. Drillon - 52
2. Apps - 50
3. Thompson - 44
...benchmark = Thompson: 44

1938-39:
2. Schriner - 44

1939-40:
2. Dumart/Bauer - 43

1940-41:
2. [various players] - 44

1941-42:
2. Patrick - 54

1942-43:
1. D. Bentley - 73
2. Cowley - 72
3. M. Bentley - 70
4. Patrick - 61
5. Taylor - 60
5. Carr - 60
...average as benchmark: 66 :: (72 - wartime fudge)

1943-44:
2. D. Bentley - 77 :: (95 - wartime fudge)

1944-45:
1. Lach - 80
2. M. Richard - 73
3. Blake - 67
4. Cowley - 65
5. Kennedy - 54
5. DeMarco - 54
5. Carveth - 54
5. Mosienko - 54
...average as benchmark: 63 :: (78 wartime fudge)

1945-46:
2. G. Stewart - 52 :: (60 wartime fudge)

1946-47:
1. M. Bentley - 72
2. M. Richard - 71
3. Taylor - 63
...benchmark = Taylor: 63

1947-48:
2. O'Connor - 60

1948-49:*
1. R. Conacher - 68
2. D. Bentley - 66
3. Abel - 54
3. Lindsay - 54
...benchmark = Abel/Lindsay: 54

1949-50:
2. Abel - 69

1950-51:
2. M. Richard - 66

1951-52:
2. Lindsay - 69

1952-53:
1. Howe - 95
2. Lindsay - 71
3. M. Richard - 61
...benchmark = M. Richard: 61

1953-54:
1. Howe - 81
2. M. Richard - 67
3. Lindsay - 62
4. Geoffrion - 54
5. Olmstead - 52
6. Kelly - 49
...average as benchmark: 61

1954-55:
2. M. Richard - 74

1955-56:
1. Beliveau - 88
2. Howe - 79
3. M. Richard - 71
...benchmark = M. Richard: 71

1956-57:
1. Howe - 89
2. Lindsay - 85
3. Beliveau - 84
4. Bathgate - 77
5. Litzenberger - 64
...benchmark = Bathgate: 77

1957-58:
1. Moore - 84
2. H. Richard - 80
3. Bathgate - 78
4. Howe - 77
5. Horvath - 66
6. Litzenberger - 62
7. MacKell - 60
8. Delvecchio - 59
...average as benchmark: 71

1958-59:
1. Moore - 96
2. Beliveau - 91
3. Bathgate - 88
4. Howe - 78
5. Litzenberger - 77
6. Geoffrion - 66
...average as benchmark: 83

1959-60:
2. Horvath - 80

1960-61:
2. Beliveau - 90

1961-62:
1. Hull/Bathgate - 84

1962-63:
2. Bathgate - 81

1963-64:
1. Mikita - 89
2. Hull - 87
3. Beliveau - 78
...benchmark = Beliveau: 78

1964-65:
2. Ullman - 83

1965-66:
2. Rousseau/Mikita - 78

1966-67:
1. Mikita - 97
2. Hull - 80
3. Ullman - 70
...benchmark = Ullman: 70

1967-68:
2. Esposito - 84

1968-69:
2. Hull - 107

1969-70:
1. Orr - 120
2. Esposito - 99
3. Mikita - 86
...benchmark = Mikita: 86

1970-71:
1. Esposito - 152
2. Orr - 139
3. Bucyk - 116
4. Hodge - 105
5. Hull - 96
6. Ullman - 85
7. Cashman - 79
8. McKenzie - 77
9. Keon - 76
9. Beliveau - 76
...average as benchmark: 100

1971-72:
2. Orr - 117

1972-73:
2. Clarke - 104

1973-74:
1. Esposito - 145
2. Orr - 122
3. Hodge - 105
4. Cashman - 89
5. Clarke - 87
6. Martin - 86
...average as benchmark: 106

1974-75:
2. Esposito - 127

1975-76:
2. Clarke - 119

1976-77:
1. Lafleur - 136
2. Dionne - 122
3. Shutt - 105
...benchmark = Shutt: 105

1977-78:
1. Lafleur - 132
2. Trottier - 123
3. Sittler - 117
4. Lemaire - 97
5. Potvin - 94
6. Bossy - 91
...average as benchmark: 109

1978-79:
1. Trottier - 134
2. Dionne - 130
3. Lafleur - 129
4. Bossy - 126
5. MacMillan - 108
6. Chouinard - 107
7. Potvin - 101
8. Federko - 95
...average as benchmark: 116

1979-80:
1. Dionne - 137
1. Gretzky - 137
3. Lafleur - 125
4. Perreault - 106
5. Rogers - 105
6. Trottier - 104
...average as benchmark: 119

1980-81:
2. Dionne - 135

1981-82:
2. Bossy - 147

1982-83:
2. Stastny - 124

1983-84:
2. Coffey - 126

1984-85:
2. Kurri - 135

1985-86:
2. Lemieux - 141

1986-87:
2. Kurri - 108

1987-88:
1. Lemieux - 168
2. Gretzky - 149
3. Savard - 131
...benchmark = Savard: 131

1988-89:
1. Lemieux - 199
2. Gretzky - 168
3. Yzerman - 155
4. Nicholls - 150
5. Brown - 115
6. Coffey - 113
7. Mullen - 110
8. Kurri - 102
...average as benchmark: 139

1989-90:
2. Messier - 129

1990-91:
1. Gretzky - 163
2. Hull - 131
3. Oates - 115
...benchmark = Oates: 115

1991-92:
1. Lemieux - 131
2. Stevens - 123
3. Gretzky - 121
4. Hull - 109
5. Messier - 107
5. Robitaille - 107
...average as benchmark: 116

1992-93:
2. Lafontaine - 148

1993-94:
2. Fedorov - 120

1994-95:
1. Lindros/Jagr - 70

1995-96:
1. Lemieux - 161
2. Jagr - 149
3. Sakic - 120
...benchmark = Sakic: 120

1996-97:
2. Selanne - 109

1997-98:
2. Forsberg - 91

1998-99:
2. Selanne - 107

1999-00:
2. Bure - 94

2000-01:
1. Jagr - 121
2. Sakic - 118
3. Elias - 96
...benchmark = Elias: 96

2001-02:
2. Naslund - 90

2002-03:
2. Naslund - 104

2003-04:
2. Sakic/Kovalchuk - 87

2005-06:
1.Thornton - 125
2. Jagr - 123
3. Ovechkin - 106
...benchmark = Ovechkin: 106

2006-07:
2. Thornton - 114

2007-08:
2. Malkin - 106

2008-09:
2. Ovechkin - 110

2009-10:
2. Ovechkin/Crosby - 109

2010-11:
2. St. Louis - 99

2011-12:
2. Stamkos - 97

2012-13:
2. Stamkos - 57

2013-14:
2. Getzlaf - 87
*see post #133 for an adjustment to D.Bentley and R. Conacher's numbers in the 1948-49 season (benchmark unchanged).

Comments, criticisms, fact-checking and personal attacks are all welcome.


Last edited by Sturminator: 01-31-2015 at 05:51 AM. Reason: added "first gap" text to Rule #3, and added Bathgate Rule, Wartime Fudge
Sturminator is offline   Reply With Quote
Old
02-27-2013, 05:55 AM
  #2
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
A few of my own comments on these numbers:

- the identification of seasons in which there is a whole tier of outlier scorers was a very interesting exercize, in and of itself. I estimated the over/under on how many times the averaging method would be necessary post-consolidation at 12. It turned out that it was necessary 14 times based on my methodology, though two of those are war years, so I'm pretty happy with the line I set.

- when scoring is averaged in order to set a benchmark, it is a clear sign that there is a large imbalance in the league. Not surprisingly, this was only necessary once in the period between consolidation and the second world war, when the league was largely playing at a high level, and there was a good deal of competitive parity. I had an idea that this might be the case, but I was somewhat surprised at how unbalanced the mid-50's look from this perspective. The simultaneous peaking of both the Production Line and the great Habs dynasty caused a violent tiering of league scoring for a little while there. The early 70's Bruins distortion is well known, though I was a bit surprised to see how distorted the late-70's ended up being - Guy Lafleur being the common denominator here. The distortions caused by Gretzky and Lemieux in the 1988-92 timeframe end up looking fairly minor in the grand scheme of things.

- the player who most stands out as being benefitted by this analysis: Andy Bathgate. It is striking to look at where Bathgate stands in the scoring races when you realize how distorted his era really was. Bathgate's status as the only non-Hab/Red Wing in the elite scorers in a few of those seasons is extremely impressive, and I think this method of benchmarking does a good job of illuminating just how great Andy's scoring feats really were.

Sturminator is offline   Reply With Quote
Old
02-27-2013, 06:37 AM
  #3
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Number of times the following systems are used in the above to establish a scoring benchmark in the post-consolidation NHL:

1. Straight Vs2: 54
2. Vs3: 17
3. VsX Averaged: 14

I think this ends up being a fairly good algorhythm. The original Vs2 system is mostly preserved because it was mostly a quite efficient way of removing outliers. The old system ran into problems, however, when more than one outlying scorer was present in a given season. The system I've used here handles tiers of 2-5 outliers in a way which seems relatively fair to all involved. I am satisfied with it, at any rate.

There are imperfections in the system, and some seasons are difficult to evaluate in spite of the flexibility of the methodology, but hopefully if everybody can reference the same data for benchmarking purposes, we can establish some kind of coherent, consolidated system for doing percentage-based evaluations of league scoring.

Sturminator is offline   Reply With Quote
Old
02-27-2013, 07:41 AM
  #4
EagleBelfour
Registered User
 
EagleBelfour's Avatar
 
Join Date: Jun 2005
Country: Canada
Posts: 7,329
vCash: 500
Although I understood the concept in the past, I never really used it to analyse offensive players. However, this is a bullet/idiot proof process that is very to look at for future references. I may jsut start adding those .Vs2 numbers in my biographies, good job Sturm (& 70's & everyone who participate in the process to come up with those formulas, i know they've been around for a few years already around these parts).

Few questions:

- Which kind of players benefits the most from using this metric to understand their offensive production? Does this metric show just as much the offensive potential of a 3rd liner than a 1st liner?

- Is every era equal in term of how effective this metric is?

- Is this metric interesting to use for defenceman also?

- Is there a percentage that the metric lose is effectveness?

- Could anyone use this metric for other individual stats, like goalscorings or assists?

I guess most of those questions sounds simplistic, but I wanna have a perfect grasp of this new method before I start using it!

EagleBelfour is offline   Reply With Quote
Old
02-27-2013, 08:01 AM
  #5
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by EagleBelfour View Post
Although I understood the concept in the past, I never really used it to analyse offensive players. However, this is a bullet/idiot proof process that is very to look at for future references. I may jsut start adding those .Vs2 numbers in my biographies, good job Sturm (& 70's & everyone who participate in the process to come up with those formulas, i know they've been around for a few years already around these parts).
The most credit should go to BM67, who pioneered this method of evaluation. He's never cared too much for laurels, but he very much deserves them in this case.

Quote:
- Which kind of players benefits the most from using this metric to understand their offensive production? Does this metric show just as much the offensive potential of a 3rd liner than a 1st liner?
The metric is best for evaluating first-line scorers who were also on their teams' respective first unit powerplays. It is based on a raw breakdown of points, so in order for two players to be fairly evaluated against one another in this system, we must assume that they had equal opportunity to score, which is obviously not always the case. It is of very limited value for lower-line players, even including prewar guys like Cecil Dillon who did not typically skate on their teams' top lines.

Quote:
- Is every era equal in term of how effective this metric is?
I think it does a pretty good job, at this point, of allowing us to compare individual scoring across eras, which is entirely the point. For comparing players who both competed within the same era, it is really not necessary, as raw points will do. This methodology is meant to cover all eras equally, and I think it does a reasonable job of that. Keep in mind, though, that we still have to take the strength of certain eras into account - so a 90 in 1944-45 is not the same as a 90 in 2010-11, because of what we know about the composition of the league at those points in time. It is an improved, though still rough metric for comparing scoring across eras.

Quote:
- Is this metric interesting to use for defenceman also?
In my opinion, no. The role of defensemen in joining the attack has changed a great deal since the 1925-26 season, and powerplay opportunities are even more important for defenseman scoring than they are for forwards. Percentage comparisons among defensemen would yield some wild distortions if attempted across eras.

Quote:
- Is there a percentage that the metric lose is effectveness?
I generally only track numbers down to 50% of the benchmark scorer, but I can't say for sure at what point the numbers become meaningless.

Quote:
- Could anyone use this metric for other individual stats, like goalscorings or assists?
Sure, it could be used to compare secondary stats, though I think this is somewhat distorting, as imbalanced scorers will end up looking better than they really were if we only look at their area of relative strength, be it goals or assists. I prefer using points as a whole because it tends to even out the imbalances and give us a truer picture of how a player performed offensively. With that in mind, I would say that goals are somewhat more valuable than assists in the modern game, and have been since the NHL began calculating second assists, for reasons which should be obvious. If anyone wanted to put together similar data for post-consolidation goalscoring benchmarks, I wouldn't object to it, but it is a bit of work.

Sturminator is offline   Reply With Quote
Old
02-27-2013, 08:23 AM
  #6
EagleBelfour
Registered User
 
EagleBelfour's Avatar
 
Join Date: Jun 2005
Country: Canada
Posts: 7,329
vCash: 500
Thanks for the quick answer, I appreciate! So I would guess that a guy like Cecil Dillon, who was on his 2nd line for most of his career, surrounded by two-way/defensive player and playing seldomly on the PP, was a far stronger offensive player than his Vs2. would indicate? It's still a nice indicative for other players than 1st liner, as we can try to understand their situation and put their numbers into context.


If I understand how the system, it would mean that the Vs.2 of a player like Gordie Drillon would be:

73, 104, 77, 93, 100, 76, 76

OR

104, 100, 93, 77, 76, 76, 73

WWII season

EagleBelfour is offline   Reply With Quote
Old
02-27-2013, 10:43 AM
  #7
BM67
Registered User
 
BM67's Avatar
 
Join Date: Mar 2002
Location: In "The System"
Country: Canada
Posts: 4,582
vCash: 500
Assist numbers are really only necessary for early era guys where the low assist numbers make everybody look like a puck hog.

Straight Vs2 for Nels Stewart have him well ahead of John Leclair for instance. His best 5 years total are just a hair higher than Alexander Mogilny, Kirk Muller and Doug Mohns.

For defensemen you really need to have both Vs2 for the league and the defensemen only. Pierre Pilote doesn't stand out against the league, 42 career average, but he does against the other defensemen, 91 career average.

It will allow you to see how involved D were in different eras. Harvey 102% jump from league to D only average. Pilote 115%, Orr 87%, Park 70%, Bourque 67%, Larry Murphy & Paul Coffey 60%.

BM67 is offline   Reply With Quote
Old
02-27-2013, 11:37 AM
  #8
EagleBelfour
Registered User
 
EagleBelfour's Avatar
 
Join Date: Jun 2005
Country: Canada
Posts: 7,329
vCash: 500
Quote:
Originally Posted by BM67 View Post
Assist numbers are really only necessary for early era guys where the low assist numbers make everybody look like a puck hog.

Straight Vs2 for Nels Stewart have him well ahead of John Leclair for instance. His best 5 years total are just a hair higher than Alexander Mogilny, Kirk Muller and Doug Mohns.

For defensemen you really need to have both Vs2 for the league and the defensemen only. Pierre Pilote doesn't stand out against the league, 42 career average, but he does against the other defensemen, 91 career average.

It will allow you to see how involved D were in different eras. Harvey 102% jump from league to D only average. Pilote 115%, Orr 87%, Park 70%, Bourque 67%, Larry Murphy & Paul Coffey 60%.
I don't understand those two examples in the Vs2. metric context. We all know that Nels Stewart is a far superior offensive player than any of the players you mention, while Bobby Orr a much greater offensive force than Pierre Pilote.

EagleBelfour is offline   Reply With Quote
Old
02-27-2013, 12:50 PM
  #9
bluesfan94
#BackesforSelke
 
bluesfan94's Avatar
 
Join Date: Jan 2008
Location: St. Louis
Country: United States
Posts: 10,220
vCash: 500
Someone mentioned that they thought we should average the top-18 scorers for each year to find a standard. Would doing this be worth the time?

bluesfan94 is online now   Reply With Quote
Old
02-27-2013, 12:57 PM
  #10
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by EagleBelfour View Post
Thanks for the quick answer, I appreciate! So I would guess that a guy like Cecil Dillon, who was on his 2nd line for most of his career, surrounded by two-way/defensive player and playing seldomly on the PP, was a far stronger offensive player than his Vs2. would indicate? It's still a nice indicative for other players than 1st liner, as we can try to understand their situation and put their numbers into context.
I haven't run the numbers for Dillon, nor do I know the exact composition of the Rangers' powerplay at the time (it is possible that Dillon was substituted for Bun Cook on the powerplay unit), but this method will most likely underrate Cecil Dillon, yes.

Quote:
If I understand how the system, it would mean that the Vs.2 of a player like Gordie Drillon would be:

73, 104, 77, 93, 100, 76, 76

OR

104, 100, 93, 77, 76, 76, 73

WWII season
Those are the numbers I have for Drillon. They are most similar to the percentage scores of Martin St. Louis.

Martin St Louis:
108, 100, 89, 86, 78, 73, 67

Gordie Drillon:
104, 100, 93, 77, 76, 76*, 73

Sturminator is offline   Reply With Quote
Old
02-27-2013, 01:03 PM
  #11
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by bluesfan94 View Post
Someone mentioned that they thought we should average the top-18 scorers for each year to find a standard. Would doing this be worth the time?
I'm not personally a fan of this idea. The goal of this system is to set a benchmark of "star" level scoring, but to eliminate the outliers who simply lapped the field. With the possible exception of the war years, every era has legitimate stars at the top of the table, and I believe this metric does a good job of setting a reasonable baseline against which we can compare the scoring of first line forwards of any era. But not every era has relevant scorers 18 deep, and including the 18 best scorers in every season would end up making players in weak eras look much better than they really were.

Our goal here should be to compare scoring to a reasonable baseline of star performance, irrespective of era. This method does not perfectly equalize the eras, but I think it does a better job than a top-18 average would do, whatever its intuitive appeal.

Sturminator is offline   Reply With Quote
Old
02-27-2013, 01:10 PM
  #12
BM67
Registered User
 
BM67's Avatar
 
Join Date: Mar 2002
Location: In "The System"
Country: Canada
Posts: 4,582
vCash: 500
Quote:
Originally Posted by EagleBelfour View Post
I don't understand those two examples in the Vs2. metric context. We all know that Nels Stewart is a far superior offensive player than any of the players you mention, while Bobby Orr a much greater offensive force than Pierre Pilote.
For Stewart, it's just in regard to his assist totals, not overall offense.

Orr scores a yearly average of 67% Vs#2 and 126% Vs#2D, while Pilote scores 42% and 91%. Pilote's Vs2D number is 115% higher than his Vs2, while Orr's is only 87% higher.

Orr is better than Pilote offensively but D scoring really started to take off after Orr. It shows not that Pilote was better, but that the D were held back more in the 50s-60s. The 115% is not a measure of Pilote as much as it is of overall D scoring during his career.

Coffey scores 92% Vs#2D. Is Pilote almost as good as Coffey offensively? If you were just looking at Vs#2D numbers you might think so.

I've seen people say Fetisov's USSR numbers don't look that great. Knowing his Vs#2 is 47, his Vs#2D is 113, and the D to league rate is 138%, should allow you to better grasp how good his USSR numbers really are. Kasatonov is 40, 89 and 122 for comparison.

Good but not great vs the league, great vs D only, and overall D scoring was very restricted.

The USSR is another place where goal to assist ratio is distorted from the modern NHL standard.

BM67 is offline   Reply With Quote
Old
02-27-2013, 02:26 PM
  #13
TheDevilMadeMe
Global Moderator
 
TheDevilMadeMe's Avatar
 
Join Date: Aug 2006
Location: Brooklyn
Country: United States
Posts: 41,857
vCash: 500
It's an improvement, but I still see some seasons as issues:

I have mixed feelings about using 10% as a benchmark in 1927,1928, and 1934 where #2 had fewer than 40 points. Even though a 4 point difference is over 10% is it really enough to make #2 an outlier when 4 points could just be 1 or 2 flukey games?

1943 and 1945 are disasters - They are already weakened because of the war (especially 1945), and this system actually reduces the standard even further, especially in the incredibly weak 1945. I think the War is the biggest reason that Cowley and the Punch Line curbstomped the rest of the league, but this system treats them as having legitimately >100 seasons. Not sure what the solution to the war years is though.

The Chicago inflation of 1949 is pretty galling, but using #2 as a benchmarks seems reasonable for all non-Chicago players.

The late 50s are a mess. Scoring in the 50s in general is a mess - it's the only time period when scoring actually increased in the playoffs, rather than decreased, and I think it's because the two best teams (Montreal and Detroit) were so much more talented than everyone else, that everyone else just lowered the average in the regular season.

I actually think the system does a great job of dealing with 1971.

1974 still looks fishy. The highest non Bruin scores 87 points, yet the standard uses 106. On the other hand, 106 fits much more nicely with the standards from 1973 and 1975, so I dunno.

The late 70s are messy, but I the system looks like it works for them.

The system seems to deal with Gretzky and Lemieux quite well, and in general has reasonable benchmarks for everything afterwards, as well.

So I'd say the problem areas with the new system are
  • the late 20s/early 30s when the difference only has to be 4 points to cross the 10% threshold.
  • the war years, especially 1945
  • the late 50s
  • probably 1974

1949 is as good as any system can possibly get it - the stats of Chicago players in the late 40s should have asterixes by them anyway.

Still, an improvement on the old formula.

TheDevilMadeMe is offline   Reply With Quote
Old
02-27-2013, 02:29 PM
  #14
TheDevilMadeMe
Global Moderator
 
TheDevilMadeMe's Avatar
 
Join Date: Aug 2006
Location: Brooklyn
Country: United States
Posts: 41,857
vCash: 500
Quote:
Originally Posted by Sturminator View Post
I'm not personally a fan of this idea. The goal of this system is to set a benchmark of "star" level scoring, but to eliminate the outliers who simply lapped the field. With the possible exception of the war years, every era has legitimate stars at the top of the table, and I believe this metric does a good job of setting a reasonable baseline against which we can compare the scoring of first line forwards of any era. But not every era has relevant scorers 18 deep, and including the 18 best scorers in every season would end up making players in weak eras look much better than they really were.

Our goal here should be to compare scoring to a reasonable baseline of star performance, irrespective of era. This method does not perfectly equalize the eras, but I think it does a better job than a top-18 average would do, whatever its intuitive appeal.
I brought it up using the average of the top 18, but after thinking about it more, I basically agree with you. The quality of the #2 scorer does change across time, but much less so than the quality of the 10th-18th best scorers.

Quote:
Originally Posted by Sturminator View Post

Those are the numbers I have for Drillon. They are most similar to the percentage scores of Martin St. Louis.

Martin St Louis:
108, 100, 89, 86, 78, 73, 67

Gordie Drillon:
104, 100, 93, 77, 76, 76*, 73
With the caveat of course that Drillon played in a weaker era (and in an era where goal scorers were benefited at the expense of playmakers).


Last edited by TheDevilMadeMe: 02-27-2013 at 02:34 PM.
TheDevilMadeMe is offline   Reply With Quote
Old
02-27-2013, 02:51 PM
  #15
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by TheDevilMadeMe View Post
[*]the late 20s/early 30s when the difference only has to be 4 points to cross the 10% threshold.
eh...there is no way to improve upon this, unfortunately, and while I agree with you that it may lead to distortions within individual seasons, they will tend to even out over the course of entire careers. edit: it also cannot account for the way assists were tabulated in this era, but as this was a subjective system at the discretion of the scorers, I don't see any way to normalize it objectively.

Quote:
[*]the war years, especially 1945
Obviously. The war years need a huge asterisk. Honestly, I wanted to just skip them because I agree with you that it was mostly a bunch of bush league bullsh!t, but I included them for completeness, and because it is difficult to draw hard lines on when they began and ended. I sincerely hope that nobody will take from this post that a 100+ result in 1944-45 is the same as for 1999-00. No system can fix the war years. They are simply a cluster****.

Quote:
[*]the late 50s
This system is actually a real improvement in how we look at the late 50's, imo, specifically because it improves the relative performances of players like Bathgate and Litzenberger, which is fully deserving. Detroit/Montreal players from this period should carry something of an asterisk, of course, as they clearly benefitted from linemates who were vastly superior to those on the other four teams in the league. I am open to suggestions as to how we could improve the accounting of this era.

Quote:
[*]probably 1974
One thing I like about the 1974 results in this system is that it gives Ken Hodge a score of less than 100. That can only be a good thing, right? I'm not at all convinced that the numbers are off in this case. It wasn't one of Clarke's best years, and Martin wasn't really a top of the table scorer. I think 1974 was a case of the Bruins being great and the rest of the league being pretty meh.


Last edited by Sturminator: 02-27-2013 at 03:05 PM.
Sturminator is offline   Reply With Quote
Old
02-27-2013, 02:52 PM
  #16
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by TheDevilMadeMe View Post
With the caveat of course that Drillon played in a weaker era (and in an era where goal scorers were benefited at the expense of playmakers).
Right on both points. Following these numbers slavishly would be a mistake.

Sturminator is offline   Reply With Quote
Old
02-27-2013, 02:54 PM
  #17
seventieslord
Moderator
 
seventieslord's Avatar
 
Join Date: Mar 2006
Location: Regina, SK
Country: Canada
Posts: 25,367
vCash: 500
A few notes…
*
1.*** You don’t like when I make exceptions, but you sure make a lot of them in the years prior to expansion! It’s not that I don’t agree with what you did in those early seasons, but the mantra has been “you have to use #2 all the time and stick to it” but as you can see, that’s often not what makes the most sense.
2.*** Is this what you’ve always been doing? Dreakmur said that yesterday he “thought” he was using your system but it’s easy to see now how he went wrong… seasons like 89 and 92 are pretty complex without having been explicitly defined like you now have.
3.*** I still don’t understand the steadfast refusal to account for Gretzky. The benchmarks being used throughout the 1980s (except 1987) are extremely unforgiving to a lot of players who peaked during that time. No other era had a player who was so far ahead of the pack and made teammates gravitate towards the top of the list. The benchmarks that you’re choosing to use demonstrate this, as they are disproportionately high and appear to cause that season to have a small number of 80+ players, for example. Again, you think the 80s were great and the 90s were (at best) equal, but the percentages under your system likely show it to be both better at the top, and deeper.
4.*** The system should still be vetted (like mine has been) to determine which seasons, if any, should possiblyhave a different benchmark used. I understand you’re choosing a system and adhering to it, but if we can identify seasons in which an abnormally large or small number of players reach scores of 70+, 80+, etc,* compared to the seasons around them (i.e. seasons where the talent pool was about the same) then it’s an indication that too harsh or too easy of a benchmark is being used. You’ve already breached the user intervention barrier anyway…
5.*** This can be ironed out as a minor detail, but we should figure out what to do for 1944 and 1945. My solution is to assume no one leaves to fight, and use the block of seasons before and after the war to determine what kind of point totals the best players would have (i.e. if Lach had 80, what would Bentley have had, knowing Bentley tended to score x% more/less than Lach?) and use the hypothetical #2 as the benchmark.
6.*** If you were telling EB that this could not be used to compare defensemen from one era to the other using their “league wide, all positions” vs#2 score, I 100% agree. However, a “defenseman % score” metric can easily be done (I’ve already done it and have been using it, it makes the same types of concessions that you do for earlier seasons and due to the smaller sample sizes there are larger gaps more often, and concessions need to be made more often). The results over the long term don’t seem to have the problems BM67 has been describing, but that’s probably because I was taking out more outliers (I believe I used the 10% rule but didn’t do the averaging thing). Orr leads at 1353 in his best 6 seasons, then Kelly at 1075, Coffey at 916, Pilote at 782, Harvey at 742, Shore at 732, Potvin at 706, Leetch at 697, Clancy at 695, Gadsby at 684, Bourque at 677, Lidstrom at 663, Park at 655, MacInnis at 634, Housley at 611, Gonchar at 604, Quackenbush at 589, Murphy at 584, Robinson at 563, Vadnais at 560, Larson at 533, Chelios at 533, Horton at 528, Stevens at 509, Clapper at 502 (only as a defenseman), Ramage at 426, Hajt at 201, and so on. This was not meant to be a comprehensive list, though I am sure I at least caught the top-13. The rest are just examples of where some other greats would fall. Looks like it may need a bit of tweaking, but it's a solid starting point, I think.

seventieslord is offline   Reply With Quote
Old
02-27-2013, 03:21 PM
  #18
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by seventieslord View Post
1.*** You don’t like when I make exceptions, but you sure make a lot of them in the years prior to expansion! It’s not that I don’t agree with what you did in those early seasons, but the mantra has been “you have to use #2 all the time and stick to it” but as you can see, that’s often not what makes the most sense.
That's never been my mantra. I believe it is necessary to have a system, and stick to it, but a system that is overly simplistic is often worse than having none, at all. What I cannot accept is a system in which players are arbitrarily removed because of their names or the names of their linemates.

Quote:
2.*** Is this what you’ve always been doing? Dreakmur said that yesterday he “thought” he was using your system but it’s easy to see now how he went wrong… seasons like 89 and 92 are pretty complex without having been explicitly defined like you now have.
No. Previously, I used the Vs2/Vs3 system. Basically, rules 1 and 2 above, but without rule #3. Dreak didn't really use that system, either. He used some sort of hybrid between that system and yours, with a few small mathematical errors thrown in for good measure.

Quote:
3.*** I still don’t understand the steadfast refusal to account for Gretzky. The benchmarks being used throughout the 1980s (except 1987) are extremely unforgiving to a lot of players who peaked during that time. No other era had a player who was so far ahead of the pack and made teammates gravitate towards the top of the list. The benchmarks that you’re choosing to use demonstrate this, as they are disproportionately high and appear to cause that season to have a small number of 80+ players, for example. Again, you think the 80s were great and the 90s were (at best) equal, but the percentages under your system likely show it to be both better at the top, and deeper.
The 90's likely were deeper in terms of "star" players, but they were not deeper at the top, which is precisely why the benchmarks for the 80's appear more stringent...because there were fewer "depth" stars who were closer to the top scorers in the league. Yes, that is a subjective judgment from having watched hockey as an adult in both eras. edit: Howe seems to have had a profound impact on Lindsay's scoring. I am not convinced that Gretzky or Lemieux are special in this sense. Also, both men did once make their teammates outliers in my system - Gretzky with Nicholls and Lemieux with Stevens. It's not like it never happened, but I just don't see the sense in removing Paul Coffey as a #2 when he scored two points more than the #3. Outliers are outliers, or they are not.

Quote:
4.*** The system should still be vetted (like mine has been) to determine which seasons, if any, should possiblyhave a different benchmark used. I understand you’re choosing a system and adhering to it, but if we can identify seasons in which an abnormally large or small number of players reach scores of 70+, 80+, etc,* compared to the seasons around them (i.e. seasons where the talent pool was about the same) then it’s an indication that too harsh or too easy of a benchmark is being used. You’ve already breached the user intervention barrier anyway…
If you want to critique the above results season-by-season, by all means. I would appreciate it. I am not working under the assumption that further improvements are impossible, but we're not going to see eye-to-eye on proposed improvements which depend on the names rather than the scoring of any group of players.

Quote:
5.*** This can be ironed out as a minor detail, but we should figure out what to do for 1944 and 1945. My solution is to assume no one leaves to fight, and use the block of seasons before and after the war to determine what kind of point totals the best players would have (i.e. if Lach had 80, what would Bentley have had, knowing Bentley tended to score x% more/less than Lach?) and use the hypothetical #2 as the benchmark.
I have no idea what to do with the war years, and to be perfectly honest, I don't much care. I have a fairly strong prejudice against the players who were young enough to serve but did not, and instead ran up the score in a bush league.


Last edited by Sturminator: 02-27-2013 at 03:41 PM.
Sturminator is offline   Reply With Quote
Old
02-27-2013, 11:18 PM
  #19
seventieslord
Moderator
 
seventieslord's Avatar
 
Join Date: Mar 2006
Location: Regina, SK
Country: Canada
Posts: 25,367
vCash: 500
Quote:
The 90's likely were deeper in terms of "star" players, but they were not deeper at the top, which is precisely why the benchmarks for the 80's appear more stringent...because there were fewer "depth" stars who were closer to the top scorers in the league. Yes, that is a subjective judgment from having watched hockey as an adult in both eras.
OK. then what we should see is a case of there being more 90+ players in the 80s, maybe about the same 80+ players, and then more 70+ players in the 90s. If the results don't look like that, then they don't really follow what our intuition tells us.

Quote:
Also, both men did once make their teammates outliers in my system - Gretzky with Nicholls and Lemieux with Stevens. It's not like it never happened, but I just don't see the sense in removing Paul Coffey as a #2 when he scored two points more than the #3. Outliers are outliers, or they are not.
The point that I apparently haven't been very good at explaining is, you shouldn't have to have an outlier-worthy point total in order to be removed. If Gretzky or Lemieux put a player in the top-3 when they clearly wouldn't have been there without them... then using them as the benchmark makes the benchmark that many more points higher than it should be.

Quote:
If you want to critique the above results season-by-season, by all means. I would appreciate it. I am not working under the assumption that further improvements are impossible, but we're not going to see eye-to-eye on proposed improvements which depend on the names rather than the scoring of any group of players.
Nothing looks particularly unreasonable until the early 70s and the all of the 80s. Once this system is "vetted" I am pretty sure we'll have a case of way too few players scoring above certain thresholds and then I would imagine you'd be open to using different numbers as benchmarks.

Quote:
I have no idea what to do with the war years, and to be perfectly honest, I don't much care. I have a fairly strong prejudice against the players who were young enough to serve but did not, and instead ran up the score in a bush league.
Yeah.... well, Cowley's points are still worth something... it's just a question of what. I don't much care for giving him a score of 105% or whatever with an asterisk, nor do I care for striking it from the records as though it didn't happen.

seventieslord is offline   Reply With Quote
Old
02-28-2013, 01:23 AM
  #20
bluesfan94
#BackesforSelke
 
bluesfan94's Avatar
 
Join Date: Jan 2008
Location: St. Louis
Country: United States
Posts: 10,220
vCash: 500
Alright I went ahead and did the averages for the top-18 scorers since the first lockout, mainly because I was bored in class when I started and once I started I had to stop at some benchmark

2011-12: 82.22
2010-11: 82.78
2009-10: 90.83
2008-09: 89.94
2007-08: 90.06
2006-07: 97.39
2005-06: 98.06
2003-04: 79.44
2002-03: 88.83
2001-02: 78.61
2000-01: 91.83
1999-00: 81.33
1998-99: 89.28
1997-98: 82.28
1996-97: 91.89
1995-96: 111.17
1994-95: 56.11

bluesfan94 is online now   Reply With Quote
Old
02-28-2013, 01:32 AM
  #21
TheDevilMadeMe
Global Moderator
 
TheDevilMadeMe's Avatar
 
Join Date: Aug 2006
Location: Brooklyn
Country: United States
Posts: 41,857
vCash: 500
Quote:
Originally Posted by bluesfan94 View Post
Alright I went ahead and did the averages for the top-18 scorers since the first lockout, mainly because I was bored in class when I started and once I started I had to stop at some benchmark

2011-12: 82.22
2010-11: 82.78
2009-10: 90.83
2008-09: 89.94
2007-08: 90.06
2006-07: 97.39
2005-06: 98.06
2003-04: 79.44
2002-03: 88.83
2001-02: 78.61
2000-01: 91.83
1999-00: 81.33
1998-99: 89.28
1997-98: 82.28
1996-97: 91.89
1995-96: 111.17
1994-95: 56.11
I think it's probably a better benchmark to use than VS2 for any post-expansion season and probably any season since the mid 50s. But during the 40s and early 50s when the talent pool was weaker, you're going to see the few true offensive stars have ridiculously high scores if you compare them to the average of the top 18.

TheDevilMadeMe is offline   Reply With Quote
Old
02-28-2013, 01:34 AM
  #22
bluesfan94
#BackesforSelke
 
bluesfan94's Avatar
 
Join Date: Jan 2008
Location: St. Louis
Country: United States
Posts: 10,220
vCash: 500
Quote:
Originally Posted by TheDevilMadeMe View Post
I think it's probably a better benchmark to use than VS2 for any post-expansion season and probably any season since the mid 50s. But during the 40s and early 50s when the talent pool was weaker, you're going to see the few true offensive stars have ridiculously high scores if you compare them to the average of the top 18.
I can go back all the way to the expansion if you guys would like.

bluesfan94 is online now   Reply With Quote
Old
02-28-2013, 02:07 AM
  #23
Sturminator
I voted for Kodos
 
Sturminator's Avatar
 
Join Date: Feb 2002
Location: West Egg, New York
Country: Ukraine
Posts: 7,547
vCash: 500
Quote:
Originally Posted by bluesfan94 View Post
I can go back all the way to the expansion if you guys would like.
Go for it, man. I'm still not sure what to do with 1995-96, but I'm very curious to see what you come up with as you go further back in history. My guess is that it starts to thin out in the seventies and the benchmarks drop somewhat...but we shall see!

Sturminator is offline   Reply With Quote
Old
02-28-2013, 05:10 AM
  #24
bluesfan94
#BackesforSelke
 
bluesfan94's Avatar
 
Join Date: Jan 2008
Location: St. Louis
Country: United States
Posts: 10,220
vCash: 500
I'm letting my professor know it's all your faults when I'm exhausted tomorrow in class.

2011-12: 82.22
2010-11: 82.78
2009-10: 90.83
2008-09: 89.94
2007-08: 90.06
2006-07: 97.39
2005-06: 98.06
2003-04: 79.44
2002-03: 88.83
2001-02: 78.61
2000-01: 91.83
1999-00: 81.33
1998-99: 89.28
1997-98: 82.28
1996-97: 91.89
1995-96: 111.17
1994-95: 56.11
1993-94: 102.83
1992-93: 123.17
1991-92: 102.89
1990-91: 105.00
1989-90: 108.06
1988-89: 113.50
1987-88: 110.22
1986-87: 100.06
1985-86: 116.50
1984-85: 114.00
1983-84: 111.89
1982-83: 108.28
1981-82: 117.22
1980-81: 109.22
1979-80: 102.22
1978-79: 99.72
1977-78: 92.83
1976-77: 93.00
1975-76: 101.28
1974-75: 100.89
1973-74: 88.61
1972-73: 93.78
1971-72: 89.11
1970-71: 87.89
1969-70: 75.61
1968-69: 84.28
1967-68: 70.50

bluesfan94 is online now   Reply With Quote
Old
02-28-2013, 05:21 AM
  #25
TheDevilMadeMe
Global Moderator
 
TheDevilMadeMe's Avatar
 
Join Date: Aug 2006
Location: Brooklyn
Country: United States
Posts: 41,857
vCash: 500
Not surprised to see 1992-93 as the highest benchmark - it really should be - that year was really wacked. I knew 1995-96 was a year where scoring spiked because the NHL had a major crackdown on obstruction during the regular season (only to forget about it entirely come playoffs), but it still surprises me to see it so closely resemble a typical season in the 1980s for scoring among top 18 players.

TheDevilMadeMe is offline   Reply With Quote
Reply

Forum Jump


Bookmarks

Thread Tools

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off



All times are GMT -5. The time now is 07:37 AM.

monitoring_string = "e4251c93e2ba248d29da988d93bf5144"
Contact Us - HFBoards - Archive - Privacy Statement - Terms of Use - Advertise - Top - AdChoices

vBulletin Copyright ©2000 - 2015, Jelsoft Enterprises Ltd.
HFBoards.com is a property of CraveOnline Media, LLC, an Evolve Media, LLC company. ©2015 All Rights Reserved.